A high spatial resolution mountainous surface temperature remote sensing inversion hybrid algorithm

Abstract

The application discloses a high spatial resolution mountainous area surface temperature remote sensing inversion hybrid algorithm and belongs to the field of thermal infrared quantitative remote sensing research; the method comprises the following steps: constructing a mountainous area thermal infrared radiation transmission model, deducing a definition formula of a mountainous area small-scale self-heating parameter, defining a mountainous area effective emission rate, and constructing a mountainous area three-dimensional thermal infrared radiation transmission model; using a split window algorithm and simulated on-board brightness temperature data of each thermal infrared band, the optimal combination of the off-ground radiation brightness temperature of each band is estimated through optimal configuration of band combination selection; using a temperature-emission rate separation algorithm, the mountainous area surface temperature and the emission rate are separated from the off-ground radiation brightness temperature of each band at the same time; the application constructs a new mountainous area three-dimensional thermal infrared radiation transmission model and a remote sensing inversion hybrid algorithm capable of accurately inverting the high spatial resolution mountainous area surface temperature, and improves the remote sensing inversion precision of the high spatial resolution mountainous area surface temperature.

Description

Technical Field

[0001] This invention relates to the field of thermal infrared quantitative remote sensing, specifically to a hybrid algorithm for remote sensing inversion of surface temperature in mountainous areas with high spatial resolution. Background Technology

[0002] Land surface temperature (LST) is a key physical quantity in the Earth's surface system, reflecting the combined effects of interactions between land, ocean, and atmosphere. It plays an indispensable role in studying the exchange of matter and energy between the Earth's surface and atmosphere, global ocean circulation, climate anomalies, resource and environmental monitoring, and the urban heat island effect, and is widely applied in many basic disciplines and major application fields. Furthermore, LST is an important land surface parameter among the fundamental climate variables identified by the United Nations Framework Convention on Climate Change and the World Meteorological Organization, impacting numerous global social challenges, including the United Nations Sustainable Development Goals (SDGs).

[0003] Furthermore, the complex geometry of mountainous areas significantly impacts thermal radiation at the high spatial resolution pixel scale; however, mountains cover approximately 24% of the global land area. Therefore, accurately retrieving mountain surface temperatures is crucial for studying both local and global climate change. High-resolution mountain surface temperatures are vital for remote sensing applications such as ecological conservation, climate change research, and evapotranspiration estimation in mountainous regions.

[0004] As is well known, existing thermal infrared (TIR) ​​land surface temperature remote sensing inversion methods are relatively mature. On flat surfaces with relatively uniform underlying surfaces, the temperature inversion error can be within 1 K. However, in areas with complex terrain, the inversion error can sometimes exceed 4 K. This is mainly because existing land surface temperature inversion methods primarily focus on uniform, flat surfaces and are based on a planar parallel thermal radiation transfer equation. Mountainous terrain is complex, and the transmission of pixel-scale thermal radiation is more complex. Compared to flat surfaces, the geometry of mountainous areas causes changes in the radiative energy received between the surface and the sensor. Although the influence of terrain geometry can be ignored in coarse-resolution TIR images, its impact is significant for high spatial resolution remote sensing images. Therefore, directly applying land surface temperature inversion algorithms based on the flat surface assumption to high spatial resolution TIR images may introduce large inversion errors.

[0005] However, the spatial resolution of TIR data from current thermal infrared sensors is mostly at the hundred-meter level or higher, making existing LST remote sensing inversion algorithms difficult to apply to high spatial resolution TIR satellite remote sensing data with geometric structures and proximity effects. Furthermore, current high spatial resolution mountain land surface temperature (MLST) inversion algorithms do not fully consider the multiple scattering effects of terrain three-dimensional structure on thermal radiation and the thermal radiation contribution of neighboring pixels. This results in large inversion errors for mountain land surface temperature, especially within pixels with large terrain undulations. Therefore, accurate inversion of high spatial resolution MLST requires the development of new mountain thermal infrared radiation transfer models to characterize the influence of the three-dimensional structure of mountain pixels on the thermal radiation transfer process, and the construction of new mountain land surface temperature inversion algorithms based on these models.

[0006] In summary, revealing the influence mechanism of the complex three-dimensional structure of mountainous terrain on thermal radiation transfer and quantifying the thermal radiation patterns of neighboring pixels are crucial for accurately retrieving high spatial resolution mountain surface temperatures. Constructing a scientifically sound method for retrieving mountain surface temperatures using thermal infrared remote sensing and developing a high-precision modeling theory suitable for diurnal scaling of mountain surface temperatures are not only core tasks of satellite thermal infrared remote sensing technology in mountainous thermal environment research but also key to a deeper understanding of mountainous climate and environmental issues. Acquiring high spatial resolution MLSTs can enhance the monitoring capabilities of mountainous thermal environments, enabling them to play a significant role in climate change research, thermal anomaly monitoring, and surface energy balance analysis, possessing profound practical significance and broad application prospects. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention provides a hybrid algorithm for remote sensing inversion of surface temperature in mountainous areas with high spatial resolution, thereby achieving high-precision remote sensing inversion of surface temperature in mountainous areas with high spatial resolution.

[0008] To achieve the above technology, the specific steps are as follows:

[0009] S1. Based on high-resolution digital elevation model (DEM) data, the small-scale self-heating parameter (SSP) of mountainous areas is derived and the heat radiation interception ratio inside the pixel is quantified. By calculating the small-scale self-heating parameter (SSP) and the emissivity of the original material, the effective emissivity of mountainous areas is defined and the expression of the effective emissivity (MLSE) of mountainous areas is derived.

[0010] In this invention, the small-scale self-heating parameters are calculated based on high-resolution digital elevation model (DEM) data, and the effective emissivity in mountainous areas is calculated using the original material emissivity of the pixels and SSP.

[0011] Specifically, this invention takes into account the influence of mountainous terrain and derives an estimation formula for small-scale self-heating parameters applicable to mountainous areas, as shown in the following expression:

[0012]

[0013] In the formula, It is the surface area projected onto the corresponding slope. It is the actual surface area of ​​the pixel. It is the pixel size of the coarse-resolution data. It refers to the pixel size of fine-resolution data. l This indicates the number of fine-resolution data pixels contained within a single coarse-resolution data pixel. g yes l The summation index, The slope angle for coarse-resolution data. SSP represents the slope angle for fine-resolution data; the value range of SSP is 0.0~1.0; SSP = 0.0 indicates that the interior of the pixel is a completely flat, ideal surface without any microscopic unevenness, therefore, it does not intercept thermal radiation from its own structure, and the radiation exchange is similar to that of a flat surface; conversely, SSP = 1.0 represents that the pixel is a completely closed box-like structure, whose internal surface can fully intercept and absorb thermal radiation from other surfaces, exhibiting radiation characteristics similar to a blackbody. In actual terrain, the SSP value reflects the complexity of the terrain structure inside the pixel; the larger the value, the more rugged the terrain, the stronger the radiation interaction between surfaces, and the more significant the impact on the balance of surface radiation energy.

[0014] Based on SSP in mountainous areas, the effective emissivity in mountainous areas is redefined; Kirchhoff's law states that absorptivity and emissivity are equal, therefore the pixel... i The first emission of radiation It can be represented as:

[0015]

[0016] In the formula, e (Ω) is the raw material emissivity (FLSE) of pixel Ω without considering terrain. B ( T s (Ω) represents the surface physical temperature of pixel Ω. T s Planck blackbody radiation at (Ω);

[0017] After the first reflection, the second radiative exitance of pixel Ω It can be represented as:

[0018]

[0019] In the formula, For pixel Ω, the small-scale self-heating parameter in mountainous areas;

[0020] After the first n After the second reflection, the radiant exitance of the pixel It can be represented as:

[0021]

[0022] The total radiative exitance of the target pixel It can be represented as:

[0023]

[0024] In the formula, n Let be the number of reflections, when n As it approaches infinity, [1-(1- e (Ω))SSP(Ω)] n The total radiative exitance of the target pixel is close to 0. This can be further expressed as:

[0025]

[0026] Based on the definition of emissivity, the redefined expression for Mountainous Effective Emissivity (MLSE) is as follows:

[0027]

[0028] In the formula, eff m (Ω) is the effective surface emissivity in mountainous areas;

[0029] As can be seen from the MLSE estimation formula, the emissivity of the target pixel increases due to multiple scattering within the pixel in mountainous areas; and the effective emissivity in mountainous areas decreases due to the influence of multiple scattering within the pixel. Emissivity higher than that of the original material e (Ω), the larger the SSP of a pixel in a mountainous area, the stronger the multiple scattering effect within the pixel, making... Closer to 1.0; when SSP(Ω) = 1.0, =1.0, at which point it resembles a blackbody; when SSP(Ω)=0.0, = e (Ω).

[0030] S2. Based on sky visibility factors, small-scale natural parameters and effective emissivity of mountainous areas, a three-dimensional thermal infrared radiation transmission (MMS-TRT) model for mountainous areas is developed and constructed by considering multiple scattering within pixels, topographic thermal radiation contribution between pixels and thermal radiation contribution of neighboring pixels, and combining physical laws and mathematical derivations.

[0031] The three-dimensional thermal infrared radiation transmission model of the mountainous area includes: the emitted radiation of the mountainous pixel itself, the multiple scattering within the pixel, the total radiance leaving the surface of the mountainous pixel, and the total thermal radiation of the mountainous pixel.

[0032] S3. Based on SSP, SVF and the mountainous three-dimensional thermal infrared radiation transfer MMS-TRT model, atmospheric parameters are simulated using 98 atmospheric profile data under selected clear sky conditions and the atmospheric radiation transfer model MODTRAN to generate a thermal infrared band atmospheric top brightness temperature dataset.

[0033] Specifically, MODTRAN 5.2 simulates atmospheric parameters by combining 98 atmospheric profile data selected from the TIGR atmospheric profile database, and then uses 82 emission spectra, topographic factors, and the developed mountainous three-dimensional thermal infrared radiation transfer MMS-TRT model to simulate the on-board brightness temperature data of the ASTER sensor in five thermal infrared (TIR) ​​bands.

[0034] In this invention, the TIGR 2000 v1.1 atmospheric profile database is used to characterize atmospheric profiles. This database contains 2311 atmospheric profiles, each comprising 40 pressure layers, ranging from 1013 hPa to 0.05 hPa. To ensure that the atmospheric profiles represent clear-sky conditions, the TIGR atmospheric profiles are filtered as follows: 1) If the relative humidity value is greater than 90% or the relative humidity of two adjacent layers is greater than 85%, the profile is deleted; 2) Profiles with a relative humidity greater than 80% at an altitude of 2 km are also removed. 98 clear-sky atmospheric profiles were selected from the TIGR 2000 v1.1 atmospheric profile database; the selected profiles can describe a wide range of clear-sky atmospheric conditions, with a bottom air temperature (…). T 0) The range is 236.25 K to 311.95 K, and the WVC range is 0.09 g / cm² to 6.15 g / cm², with 6 uniformly distributed WVC ranges: 0-1, 1-2, 2-3, 3-4, 4-5, and greater than 5 g / cm²; 3) The bottom temperature T0 of the atmospheric profile is perturbed and set to [T0 -10 K, T0 + 20 K] with a step size of 5 K, resulting in 7 cases; 4) In order to better characterize the terrain undulations within and between pixels in mountainous areas, both SSP and SVF are set to 0.0~1.0 with a step size of 0.1, resulting in 11 values ​​for both SSP and SVF; The on-board brightness temperature data of the five thermal infrared bands of the ASTER sensor were simulated.

[0035] S4. Based on the atmospheric top brightness temperature dataset of the thermal infrared band, the split window algorithm is improved by precise atmospheric correction and classification combination considering the small-scale self-heating parameter SSP of terrain factor and the sky visibility factor SVF. The optimal channel combination for estimating the ground-level radiation brightness temperature of each thermal infrared band is constructed, and the improved split window I-SW algorithm is obtained. Then, a split window coefficient lookup table for mountainous areas is constructed.

[0036] Specifically, this invention treats the regression coefficients of the split-window SW algorithm as functions of the observed celestial angles VZA, SSP, and SVF, thereby improving the accuracy of the SW algorithm in estimating the ground-based radiative brightness temperature (LGBT) in complex mountainous terrain. The improved split-window I-SW algorithm expression is as follows:

[0037] In the formula, This represents the brightness temperature of the ground-based radiation in the thermal infrared band. SSP and SVF are the small-scale autothermal parameter and the spectrophotometric factor, respectively. A 0~ A 4 represents the regression coefficients of the I-SW algorithm. The first regression coefficient, for The regression coefficient, The third regression coefficient, The fourth regression coefficient, The fifth regression coefficient, To observe the zenith angle, It is the thermal infrared band i The brightness of the star, It is the thermal infrared band j The brightness temperature of the star. Because ASTER's VZA is relatively small (±8.5)... ◦ This invention uses regression coefficients that are independent of VZA;

[0038] This invention combines the selected TIGR clear-sky atmospheric profile and MODTRAN 5.2 to obtain atmospheric parameters, and then uses the MMS-TRT model to perform a forward simulation of the on-board brightness temperature data of five TIR bands of ASTER. Finally, the I-SW coefficients are determined by regression fitting.

[0039] S5. Based on SSP and effective reflectance in mountainous areas (MLSE), and considering topography and proximity effects, the empirical relationship between the minimum emissivity and the difference between minimum and maximum emissivity in the temperature-emissivity separation algorithm is recalibrated. e min The regression coefficients of MMD are used to obtain the temperature-emissivity separation algorithm for mountainous areas. e min -Regression coefficients of the empirical relationship between MMD;

[0040] Specifically, the mountainous terrain causes multiple scattering of thermal radiation within the pixel, altering the effective emissivity of the pixel. Therefore, it is necessary to redetermine the effective emissivity by combining the pixel's SSP (Surface Spatial Scattering). e min -Regression coefficients of the empirical relationship between MMD; This invention recalculates the effective emissivity in MLSE based on the formula for calculating effective emissivity in mountainous areas, combined with 82 spectral emissivity curves selected from the SSP and ECOSTRESS spectral libraries, and redetermines the corresponding values ​​for different SSPs. e min - Regression coefficients of the empirical relationship between MMD; where SSP ranges from 0.0 to 1.0 with a step size of 0.1;

[0041] S6. Based on the MMS-TRT model, split window coefficient lookup table, SSP and SVF, the expression for mountain surface temperature MLST is derived. The mountain surface temperature MLST and mountain effective reflectance MLSE are inverted by the mountain temperature-emissivity TES algorithm and iteratively looped. Finally, the inversion results of mountain surface temperature and emissivity considering topography and proximity effects are output, and the high spatial resolution hybrid remote sensing inversion algorithm for mountain surface temperature is completed.

[0042] The iterative loop is performed as follows: iteration stops when the output difference of the continuous surface temperature is less than 0.1 K or when 12 iterations are completed;

[0043] MLST can ultimately be represented as:

[0044]

[0045] In the formula, b max This indicates the band corresponding to the maximum emissivity in the thermal infrared band. Indicates band b max The inverse of the Planck function, e max The maximum transmit rate for all channels. R terrain，bmax express b max The contribution of reflected thermal radiation from adjacent terrain, Indicates band b max Downward atmospheric radiation Indicates band b max Planck function, SVF represents the sky visibility factor, T b，bmaxThis represents the ground-level radiation temperature corresponding to the band with the highest emissivity in the thermal infrared band. The mountainous TES algorithm uses an iterative method to separate MLST and MLSE. The termination criterion is that the difference between the LST outputs of two adjacent iterations is less than 0.1 K or 12 iterations are completed. When the iteration termination condition is met, the mountainous surface temperature MLST and the mountainous effective reflectivity MLSE can be separated.

[0046] Therefore, the present invention employs the above-mentioned high spatial resolution mountainous area surface temperature hybrid remote sensing inversion algorithm, which has the following beneficial effects:

[0047] (1) This invention elucidates the thermal radiation transmission process under the influence of the complex three-dimensional structure features of mountainous terrain, quantifies the effects of terrain and proximity effects, and constructs a three-dimensional thermal infrared radiation transmission model for mountainous areas. The three-dimensional structure of mountainous pixels will enhance the multiple scattering of thermal radiation within the pixel and the contribution of thermal radiation from surrounding pixels. This effect is not significant at the kilometer scale, but it cannot be ignored in high spatial resolution thermal infrared images. This invention develops a small-scale self-heating parameter for mountainous areas to characterize the thermal radiation interception ratio within pixels, and uses the sky visibility factor, which characterizes the three-dimensional structure features of mountainous areas, to quantify the geometric structure between target pixels and neighboring pixels. Based on considering the multiple scattering within mountainous pixels and the external terrain and proximity effects, a mountainous pixel-scale thermal radiation model is constructed, providing the necessary model foundation and simulation dataset for developing new mountainous pixel temperature inversion algorithms.

[0048] (2) This invention improves the accuracy of thermal infrared remote sensing inversion of surface temperature in mountainous areas by developing a hybrid remote sensing inversion algorithm suitable for high-resolution thermal infrared data. Currently, surface temperature remote sensing inversion algorithms for flat surfaces are well-developed, but for complex mountainous terrain, there is still no algorithm that fully considers terrain and proximity effects. This invention is based on a developed three-dimensional thermal radiation transfer model for mountainous areas, improves existing hybrid algorithms, eliminates the thermal radiation contribution of terrain and proximity effects, corrects simulation data and algorithm coefficients, and develops a hybrid remote sensing inversion algorithm suitable for surface temperature in mountainous areas, thereby improving the accuracy of high spatial resolution surface temperature inversion in mountainous areas. Attached Figure Description

[0049] Figure 1 This is a flowchart of a method for high spatial resolution mountain surface temperature remote sensing inversion hybrid algorithm according to the present invention.

[0050] Figure 2 This is a schematic diagram illustrating the empirical relationship between minimum emissivity and the difference between minimum and maximum emissivity in the high spatial resolution mountainous area surface temperature hybrid remote sensing inversion algorithm of the present invention; wherein, Figure (a) shows the empirical relationship between minimum emissivity and the difference between minimum and maximum emissivity without considering the terrain effect within the pixel. e min-The regression coefficient fitting plot of MMD, Figure (b) shows the empirical relationship between minimum emissivity and the difference between minimum and maximum emissivity considering the intra-pixel terrain effect. e min -MMD regression coefficient fitting plot. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0052] Reference Figure 1 A high spatial resolution mountain surface temperature remote sensing inversion hybrid algorithm, specifically including the following steps:

[0053] S1. Based on high-resolution digital elevation model (DEM) data, the small-scale self-heating parameter (SSP) of mountainous areas is derived and the heat radiation interception ratio inside the pixel is quantified. By calculating the small-scale self-heating parameter (SSP) and the emissivity of the original material, the effective emissivity of mountainous areas is defined and the expression of the effective emissivity (MLSE) of mountainous areas is derived.

[0054] In this invention, the small-scale self-heating parameters are calculated based on high-resolution digital elevation model (DEM) data, and the effective emissivity in mountainous areas is calculated using the original material emissivity of the pixels and SSP.

[0055] Specifically, this invention takes into account the influence of mountainous terrain and derives an estimation formula for small-scale self-heating parameters applicable to mountainous areas, as shown in the following expression:

[0056]

[0057] In the formula, It is the surface area projected onto the corresponding slope. It is the actual surface area of ​​the pixel. It is the pixel size of the coarse-resolution data. It refers to the pixel size of fine-resolution data. l This indicates the number of fine-resolution data pixels contained within a single coarse-resolution data pixel. g yes l The summation index, The slope angle for coarse-resolution data. SSP represents the slope angle for fine-resolution data; the value range of SSP is 0.0~1.0; SSP = 0.0 indicates that the interior of the pixel is a completely flat, ideal surface without any microscopic unevenness, therefore, it does not intercept thermal radiation from its own structure, and the radiation exchange is similar to that of a flat surface; conversely, SSP = 1.0 represents that the pixel is a completely closed box-like structure, whose internal surface can fully intercept and absorb thermal radiation from other surfaces, exhibiting radiation characteristics similar to a blackbody. In actual terrain, the SSP value reflects the complexity of the terrain structure inside the pixel; the larger the value, the more rugged the terrain, the stronger the radiation interaction between surfaces, and the more significant the impact on the balance of surface radiation energy.

[0058] Based on SSP in mountainous areas, the effective emissivity in mountainous areas is redefined; Kirchhoff's law states that absorptivity and emissivity are equal, therefore the pixel... i The first emission of radiation It can be represented as:

[0059]

[0060] In the formula, e (Ω) is the raw material emissivity (FLSE) of pixel Ω without considering terrain. B ( T s (Ω) represents the surface physical temperature of pixel Ω. T s Planck blackbody radiation at (Ω);

[0061] After the first reflection, the second radiative exitance of pixel Ω It can be represented as:

[0062]

[0063] In the formula, For pixel Ω, the small-scale self-heating parameter in mountainous areas;

[0064] After the first n After the second reflection, the radiant exitance of the pixel It can be represented as:

[0065]

[0066] The total radiative exitance of the target pixel It can be represented as:

[0067]

[0068] In the formula, n Let be the number of reflections, when nAs it approaches infinity, [1-(1- e (Ω))SSP(Ω)] n The total radiative exitance of the target pixel is close to 0. This can be further expressed as:

[0069]

[0070] Based on the definition of emissivity, the redefined expression for Mountainous Effective Emissivity (MLSE) is as follows:

[0071]

[0072] In the formula, It is the effective emission rate in mountainous areas;

[0073] As can be seen from the MLSE estimation formula, the emissivity of the target pixel increases due to multiple scattering within the pixel in mountainous areas; and the effective emissivity in mountainous areas decreases due to the influence of multiple scattering within the pixel. Emissivity higher than that of the original material e (Ω), the larger the SSP of a pixel in a mountainous area, the stronger the multiple scattering effect within the pixel, making... Closer to 1.0; when SSP(Ω) = 1.0, =1.0, at which point it resembles a blackbody; when SSP(Ω)=0.0, = e (Ω).

[0074] S2. Based on sky visibility factors, small-scale natural parameters and effective emissivity of mountainous areas, a three-dimensional thermal infrared radiation transmission (MMS-TRT) model for mountainous areas is developed and constructed by considering multiple scattering within pixels, topographic thermal radiation contribution between pixels and thermal radiation contribution of neighboring pixels, and combining physical laws and mathematical derivations.

[0075] The three-dimensional thermal infrared radiation transmission model of mountainous areas includes: the emitted radiation of the mountainous pixels themselves, multiple scattering within the pixels, the total radiance leaving the surface of the mountainous pixels, and the total thermal radiation of the mountainous pixels.

[0076] In this invention, considering multiple scattering within a pixel, the contribution of topographic thermal radiation between pixels, and the contribution of thermal radiation from neighboring pixels, a three-dimensional thermal infrared radiation transmission model for mountainous areas is derived by combining mathematical and physical knowledge.

[0077] Specifically, this invention assumes that pixels are homogeneous and at the same temperature, and that multiple scattering occurs within a single pixel. Combining physical and mathematical knowledge, a three-dimensional thermal infrared radiation transfer (MMS-TRT) model for multiple scattering in mountainous canopies is derived. If the three-dimensional structural effect of the terrain is considered but the thermal heterogeneity effect is ignored, the emissivity of a pixel Ω with a cavity effect (from SSP) and the emitted radiation of the mountainous pixel itself can be calculated as follows:

[0078]

[0079] In the formula, It is the effective emission rate in mountainous areas. B ( T s (Ω) represents the surface physical temperature of pixel Ω. T s Planck blackbody radiation at (Ω);

[0080] Therefore, the radiance of thermal radiation at the pixel scale leaving the Earth's surface and entering the atmosphere can be expressed as the sum of two parts:

[0081]

[0082] In the formula, R emit The radiation emitted by the pixels themselves in mountainous areas; R multi It is the radiation that is scattered multiple times within a pixel due to the three-dimensional structure of the terrain in mountainous areas, and it consists of three components: ① R pixel The increased thermal radiation is due to the multiple scattering of the pixel's own thermal radiation within the pixel; ② R atm ③ Atmospheric downward radiation reflected by the target pixel; R terrain The contribution of thermal radiation from the adjacent terrain reflected by the target pixel; R multi It can be represented as:

[0083]

[0084] Sky Visibility Factor (SVF) is defined as the proportion of thermal radiation a pixel can receive from hemispherical space, typically used to characterize the impact of terrain on radiation exchange of a target pixel. In mountainous environments, SVF is constrained by the shading effect of surrounding terrain; a lower SVF represents stronger terrain shading, thus affecting surface energy balance and LST retrieval accuracy. More importantly, when the search radius for estimating SVF exceeds 1.5 km, its value no longer changes significantly (i.e., tends to stabilize), indicating that further expanding the search range at a larger scale does not significantly affect the SVF calculation results. Furthermore, in the angle discretization calculation, this invention uses 16 directions for searching to improve calculation accuracy while maintaining computational efficiency. SVF is defined as the proportion of radiation a pixel can receive from hemispherical space, and can be expressed as:

[0085]

[0086] In the formula, n' represents the number of search directions. The index for the summation of the search direction numbers. The vertical elevation angle is 0.0. The SVF value ranges from 0.0 to 1.0. An SVF value close to 1.0 indicates that the target area has relatively flat terrain with almost no obstruction in the hemispherical space and extremely high visibility. This is usually found in terrain environments such as mountain tops or open plains. When the SVF value approaches 0.0, it indicates that the terrain is undulating and the surrounding features or terrain have a significant occlusion effect, resulting in very limited sky radiation that the target pixel can receive. This situation usually occurs in complex terrain areas such as deep valleys or steep canyons.

[0087] According to the definition of SSP, the size of SSP is related to the self-heating or residual thermal radiation capacity inside the pixel from a hemispherical perspective; SVF value is only related to the hemispherical perspective; furthermore, they are not affected by the observation angle of the satellite sensor; to describe in detail the derivation process of the MMS-TRT model, this invention assumes a mountainous planar pixel. i thermal radiation R 0(Ω) = ε(Ω)B( T s (Ω)); Therefore, after the first scattering of the above three radiation components, the pixel i thermal radiation It can be represented as:

[0088]

[0089] In the formula, (1-ε(Ω)) represents the reflectivity of the mountain canopy of pixel Ω, and (1-SSP(Ω)) represents the amount of heat radiation transmitted from pixel Ω to the top of the mountain canopy. This represents the downward atmospheric radiation reflected by the target pixel; R terrain The radiance represents the contribution of thermal radiation from the surrounding terrain reflected by the target pixel; (1-SVF(Ω)) represents the proportion of thermal radiation from the surrounding terrain that can enter the target pixel, which can also be expressed as TCF, called the terrain configuration factor; then, after the second scattering of pixel Ω, the radiance leaving the pixel surface... It can be represented as:

[0090]

[0091] Similarly, scattering at pixel Ω n After that, leave the pixel surface radiance It can be represented as:

[0092]

[0093] After accumulation, the solution for intra-pixel multiple scattering can be obtained. , can be represented as:

[0094]

[0095] Therefore, the total radiance of pixel Ω after multiple scatterings, leaving the mountain canopy and reaching the sky, can be expressed as:

[0096]

[0097] In the formula, Therefore, after multiple scatterings within the pixel and contributions from topographic thermal radiation between pixels, the total radiance ultimately leaving the surface of the pixel in the mountainous area is... The expression is:

[0098]

[0099] Next, in order to simplify the expression, the present invention sets e m (Ω)= +(1- )) SSP (Ω) e (Ω), The effective emissivity (MLSE) expression for mountainous areas is (1- The effective reflectance of the mountain canopy is denoted as ). SSP (Ω) e (Ω) represents the proportion of radiation blocked by the mountainous terrain geometry within the pixel; therefore, (1- ) SSP (Ω) e (Ω) is a pixel i The radiating portion between the inner patches, and the proportion of pixels leaving the mountainous area and reaching the sky; R canopy (Ω) can be simplified to:

[0100]

[0101] In summary, after obtaining the canopy thermal radiation of mountain pixels, the contribution of thermal radiation from neighboring pixels into the field of view of the target pixel is considered. R dif Along with atmospheric upward radiation, the radiation enters the satellite's thermal infrared sensor. Therefore, the total thermal radiation received by the sensor from mountainous pixels can be expressed as:

[0102]

[0103] In the formula, It is atmospheric upward radiation. It is downward atmospheric radiation. t dir It is the direct atmospheric transmittance. t dif It is atmospheric diffuse transmittance. This is atmospheric diffuse radiation. BT m It is the brightness temperature at the top of the atmosphere. B ( T s The surface physical temperature is T s Planck blackbody radiation at that time It is the radiating portion between patches within the target pixel, and the proportion of pixels leaving the mountainous area that reach the sky, B denoted by Planck's function; when SSP=0.0 and SVF=1.0, the MMS-TRT model becomes the traditional thermal infrared radiation transfer model;

[0104] S3. Based on SSP, SVF and the mountainous three-dimensional thermal infrared radiation transfer MMS-TRT model, atmospheric parameters are simulated using 98 atmospheric profile data under selected clear sky conditions and the atmospheric radiation transfer model MODTRAN to generate a thermal infrared band atmospheric top brightness temperature dataset.

[0105] The thermal infrared bands are the five thermal infrared bands of the ASTER sensor.

[0106] Specifically, MODTRAN 5.2 simulates atmospheric parameters by combining 98 atmospheric profile data selected from the TIGR atmospheric profile database, and then uses 82 emission spectra, topographic factors, and the developed mountainous three-dimensional thermal infrared radiation transfer MMS-TRT model to simulate the on-board brightness temperature data of the ASTER sensor in five thermal infrared (TIR) ​​bands.

[0107] In this invention, the TIGR 2000 v1.1 atmospheric profile database is used to characterize atmospheric profiles. This database contains 2311 atmospheric profiles, each comprising 40 pressure layers, ranging from 1013 hPa to 0.05 hPa. To ensure that the atmospheric profiles represent clear-sky conditions, the TIGR atmospheric profiles are filtered as follows: 1) If the relative humidity value is greater than 90% or the relative humidity of two adjacent layers is greater than 85%, the profile is deleted; 2) Profiles with a relative humidity greater than 80% at an altitude of 2 km are also removed. 98 clear-sky atmospheric profiles were selected from the TIGR 2000 v1.1 atmospheric profile database; the selected profiles can describe a wide range of clear-sky atmospheric conditions, with a bottom air temperature (…). T0) The range is 236.25 K to 311.95 K, and the WVC range is 0.09 g / cm² to 6.15 g / cm², with 6 uniformly distributed WVC ranges: 0-1, 1-2, 2-3, 3-4, 4-5, and greater than 5 g / cm²; 3) The bottom temperature T0 of the atmospheric profile is perturbed and set to [T0 -10 K, T0 + 20 K] with a step size of 5 K, resulting in 7 cases; 4) In order to better characterize the terrain undulations within and between pixels in mountainous areas, both SSP and SVF are set to 0.0~1.0 with a step size of 0.1, resulting in 11 values ​​for both SSP and SVF; The on-board brightness temperature data of the five thermal infrared bands of the ASTER sensor were simulated.

[0108] S4. Based on the atmospheric top brightness temperature dataset of the thermal infrared band, the split window algorithm is improved by precise atmospheric correction and classification combination considering the small-scale self-heating parameter SSP of terrain factor and the sky visibility factor SVF. The optimal channel combination for estimating the ground radiation brightness temperature of each thermal infrared band is constructed. After obtaining the improved split window algorithm, a split window coefficient lookup table for mountainous areas is constructed.

[0109] Specifically, this invention treats the regression coefficients of the split-window SW algorithm as functions of the observed celestial angles VZA, SSP, and SVF, thereby improving the accuracy of the SW algorithm in estimating the ground-based radiative brightness temperature (LGBT) in complex mountainous terrain. The improved split-window I-SW algorithm expression is as follows:

[0110]

[0111] In the formula, This represents the brightness temperature of the ground-based radiation in the thermal infrared band. SSP and SVF are the small-scale autothermal parameter and the spectrophotometric factor, respectively. A 0~ A 4 represents the regression coefficients of the I-SW algorithm. The first regression coefficient, The second regression coefficient, The third regression coefficient, The fourth regression coefficient, The fifth regression coefficient, To observe the zenith angle, It is the thermal infrared band i Brightness of the star It is the thermal infrared band j The on-board brightness temperature; due to ASTER's relatively small VZA (±8.5) ◦ This invention uses regression coefficients that are independent of VZA;

[0112] This invention combines selected TIGR clear-sky atmospheric profiles and MODTRAN 5.2 to obtain atmospheric parameters, and then uses the MMS-TRT model to positively simulate the on-board brightness temperature data of the five TIR bands of ASTER. The I-SW coefficients are then determined through regression fitting. This invention considers the small observation zenith angle of the large ASTER TIR sensor and assumes that the sensor is observing vertically. Therefore, only the observation case with VZA=0° needs to be simulated. The simulation data takes into account the following three aspects;

[0113] 1) This invention uses atmospheric parameters simulated from 98 atmospheric profile data selected from the TIGR atmospheric profile and 82 emissivity curves selected from the ECOSTRESS spectral library to simulate on-board brightness temperature data.

[0114] 2) Temperature of the lower atmosphere profile T 0 is used for perturbation, set to [ T 0-10 K, T [0+20K], with a step size of 5K, resulting in 7 possible cases;

[0115] 3) In order to better characterize the terrain undulations within and between pixels in mountainous areas, both SSP and SVF are set to 0.0~1.0 with a step size of 0.1, so both SSP and SVF have 11 values;

[0116] The simulation process assumes that neighboring pixels and the target pixel have the same LST and LSE. Based on this, and combining atmospheric transmittance, upward radiation, downward radiation, LST, LSE, SSP, and SVF, the on-board brightness temperature of ASTER's five TIR bands was simulated using the MMS-TRT model. For each band, a total of 98 atm × 7 LST × 82 LSE × 11 SSP × 11 SVF = 6,806,492 mountainous scenes were simulated. When SSP=0.0 and SVF=1.0, it represents the on-board brightness temperature dataset simulated by the traditional TIR RT model without considering the TA effect. Five regression coefficients from the ISW algorithm were fitted to all scenes, and a LUT was then constructed.

[0117] S5. Based on SSP and the effective emissivity (MLSE) in mountainous areas, and considering terrain and proximity effects, the regression coefficients of the empirical relationship between the minimum emissivity and the difference between the minimum and maximum emissivity in the temperature-emissivity separation algorithm are recalibrated to obtain the temperature-emissivity separation algorithm for mountainous areas. e min - The algorithm and regression coefficients of the MMD empirical relationship;

[0118] Specifically, mountainous terrain causes multiple scattering of thermal radiation within pixels, altering the effective emissivity of the pixels. Therefore, it is necessary to redetermine the separation algorithm based on the pixel's SSP (Surface Spatial Spread). emin -Regression coefficients of the empirical relationship between MMD and MLSE. This invention, based on the effective emissivity calculation formula for mountainous areas, combined with 82 spectral emissivity curves selected from the SSP and ECOSTRESS spectral libraries, recalculated the emissivity in MLSE and redefined the corresponding values ​​for different SSPs. e min - Regression coefficients of the empirical relationship between MMD and SSP. SSP ranges from 0.0 to 1.0 with a step size of 0.1; its good performance is as follows: Figure 2 As shown, Figure (a) illustrates the empirical relationship between minimum emissivity and the difference between minimum and maximum emissivity without considering the internal topographic effect of pixels. e min The regression coefficient fitting plot of -MMD shows the goodness of fit of the empirical relationship under different SSP values. Figure (b) shows the empirical relationship between minimum emissivity and the difference between minimum and maximum emissivity considering the intra-pixel terrain effect. e min -MMD regression coefficient fitting plot, SSP range from 0.0 to 1.0, step size 0.1.

[0119] S6. Based on the MMS-TRT model, split window coefficient lookup table, SSP and SVF, derive the MLST expression, and use the mountain temperature-emissivity TES algorithm to invert the mountain surface temperature MLST and mountain effective emissivity MLSE in an iterative loop. Finally, output the inversion results of mountain surface temperature and emissivity considering topography and proximity effects, and complete the mountain high spatial resolution surface temperature hybrid remote sensing inversion algorithm.

[0120] The iterative loop is performed as follows: the iteration stops when the output difference of the continuous surface temperature is less than 0.1 K or when 12 iterations are completed.

[0121] Specifically, the mountainous TES algorithm is similar to the traditional TES algorithm. After obtaining LGBT data using the I-SW algorithm and acquiring pixel-by-pixel atmospheric downdraft radiation using the MSTI method, the TES algorithm can be driven to retrieve surface temperature. To calculate the thermal radiation contribution of the TA effect, it is necessary to obtain LST and LSE without considering the TA effect (SSP=0.0, SVF=1.0), a process similar to separating LST and LSE in the traditional TES algorithm. Then, the LST and LSE obtained without considering the TA effect are used to calculate the thermal radiation contribution of adjacent pixels. Subsequently, 3D-LUT is used to estimate the LGBT data of mountainous pixels, and the algorithm is recalibrated under different SSPs. e minThe regression coefficients of the MMD empirical relationship are used, and then the TES algorithm is used to invert the mountain surface temperature MLST and the mountain effective emissivity MLSE. In fact, without considering the TA effect, the LST and LSE calculated by the TES algorithm are only the initial values ​​provided when calculating the thermal radiation of the terrain and adjacent pixels. Therefore, this invention mainly describes in detail the steps of inverting MLST using the TES algorithm considering the TA effect.

[0122] Based on the MMS-TRT model, surface radiation temperature in mountainous areas T ri It can be represented as:

[0123]

[0124] In the formula, B-1i For band i The inverse of the Planck function, It is a band i Downward atmospheric radiation e max The maximum emissivity across all bands. R terrain，i Indicates band i The contribution of reflected thermal radiation from adjacent terrain, the present invention sets the initial... e max =0.99; Get the value of each band T ri Then, the NEM temperature value T NEM It can be represented as:

[0125]

[0126] According to the obtained T ri and T NEM The initial surface emissivity for each TIR band can be obtained. e i for:

[0127]

[0128] Then, implement the RAT module, where the LSE ratio is:

[0129]

[0130] In the formula, N This represents the total number of bands in the TIR band. β i For band i LSE ratio;

[0131] Next, the MMD module was used to simultaneously separate LST and LSE; considering the pixel multiple scattering caused by the rugged terrain inside the mountain pixels, the emissivity of the mountain pixels was actually improved. Within the SSP range of 0.1 to 1.0, the emissivity was recalculated in increments of 0.1. e min -Regression coefficients of the MMD relationship. e min The empirical formula for MMD is:

[0132]

[0133] and

[0134]

[0135] In the formula, a The first empirical regression coefficient, b The regression coefficients of the second empirical relationship and c The third empirical regression coefficient can be obtained by using the least squares method to estimate the LSE value of the TIR band from the selected emissivity spectral curve; based on the minimum emissivity, the values ​​for each TIR band are redefined. e i :

[0136]

[0137] Based on the above derivation, MLST can ultimately be expressed as:

[0138]

[0139] In the formula, b max This indicates the band corresponding to the maximum emissivity in the thermal infrared band. Indicates band b max The inverse of the Planck function, e max The maximum transmit rate for all channels. R terrain，bmax express b max The contribution of reflected thermal radiation from adjacent terrain, Indicates band b max Downward atmospheric radiation Indicates band b max Planck function, SVF represents the sky visibility factor, T b，bmaxThis represents the ground-level radiation temperature corresponding to the band with the highest emissivity in the thermal infrared band. The mountainous TES algorithm uses an iterative method to separate MLST and MLSE. The termination criterion is that the difference between the LST outputs of two adjacent iterations is less than 0.1 K or 12 iterations are completed. When the iteration termination condition is met, the mountainous surface temperature MLST and the mountainous effective emissivity MLSE can be separated.

[0140] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A high spatial resolution mountainous area surface temperature hybrid remote sensing inversion algorithm, characterized in that, Includes the following steps: S1. Based on high-resolution digital elevation model (DEM) data, the small-scale self-heating parameter (SSP) of mountainous areas is derived and the heat radiation interception ratio inside the pixel is quantified. By calculating the small-scale self-heating parameter (SSP) and the emissivity of the original material, the effective emissivity of mountainous areas is defined and the expression of the effective emissivity (MLSE) of mountainous areas is derived. S2. Based on the sky visibility factor SVF, small-scale natural parameter SSP and mountain effective emissivity, by considering multiple scattering within pixels, topographic thermal radiation contribution between pixels and thermal radiation contribution of neighboring pixels, and combining physical laws and mathematical derivations, a three-dimensional thermal infrared radiation transmission (MMS-TRT) model for mountainous areas is developed and constructed. The three-dimensional thermal infrared radiation transmission model of the mountainous area includes: the emitted radiation of the mountainous pixel itself, the multiple scattering within the pixel, the total radiance leaving the surface of the mountainous pixel, and the total thermal radiation of the mountainous pixel. S3. Based on SSP, SVF and the mountainous three-dimensional thermal infrared radiation transfer MMS-TRT model, atmospheric parameters are simulated using atmospheric profile data under selected clear sky conditions and the atmospheric radiation transfer model MODTRAN to generate a thermal infrared band atmospheric top brightness temperature dataset. S4. Based on the atmospheric top brightness temperature dataset of the thermal infrared band, the split window algorithm is improved by precise atmospheric correction and classification combination considering the small-scale self-heating parameter SSP of terrain factor and the sky visibility factor SVF. The optimal channel combination for estimating the ground-level radiation brightness temperature of each thermal infrared band is constructed, and the improved split window I-SW algorithm is obtained. Then, a split window coefficient lookup table for mountainous areas is constructed. S5. Based on SSP and the effective emissivity (MLSE) in mountainous areas, and considering terrain and proximity effects, the regression coefficients of the empirical relationship between the minimum emissivity and the difference between the minimum and maximum emissivity in the temperature-emissivity separation algorithm are recalibrated to obtain the ε of the temperature-emissivity separation algorithm in mountainous areas. min -Regression coefficients of the empirical relationship between MMD; S6. Based on the MMS-TRT model, split window coefficient lookup table, SSP and SVF, the expression for mountain surface temperature MLST is derived. The mountain surface temperature MLST and mountain effective reflectance MLSE are inverted by the mountain temperature-emissivity TES algorithm and iteratively looped. Finally, the inversion results of mountain surface temperature and emissivity considering topography and proximity effects are output, thus completing the mountain high spatial resolution hybrid remote sensing inversion algorithm for surface temperature.

2. The high spatial resolution mountainous area surface temperature hybrid remote sensing inversion algorithm according to claim 1, characterized in that, In S1, the expression for calculating the small-scale self-heating parameter SSP in mountainous areas is: ； In the formula, It is the surface area projected onto the corresponding slope. It is the actual surface area of ​​the pixel. It is the pixel size of the coarse-resolution data. is the cell size of the fine-resolution data, l represents the number of fine-resolution data cells contained in a coarse-resolution data cell, and g is the summation index of l. The slope angle for coarse-resolution data. The slope angle for high-resolution data; The expression for the effective emissivity (MLSE) in mountainous areas is: ； In the formula, It is the effective emission rate in mountainous areas. ε(Ω) is the small-scale self-heating parameter of pixel Ω in mountainous areas, and ε(Ω) is the original material emissivity of pixel Ω without considering the terrain.

3. The high spatial resolution mountainous area surface temperature hybrid remote sensing inversion algorithm according to claim 1, characterized in that, In S2, the expression for calculating the total radiance away from the surface of the pixel in the mountainous area is: ； In the formula, let = +(1- )SSP(Ω)ε(Ω), It is the effective emissivity in mountainous areas, (1- SSP(Ω) represents the effective reflectance of the mountain canopy, ε(Ω) represents the proportion of radiation blocked by the mountainous terrain geometry within the pixel, and ε(Ω) represents the original material emissivity of pixel Ω without considering the terrain. It is the total radiance of the pixel surface away from the mountainous area. R represents the downward atmospheric radiation reflected by the target pixel; terrain (Ω) represents the thermal radiation contribution of the adjacent terrain reflected by the target pixel, Ω represents the mountain pixel, SVF represents the sky visibility factor, B(T) s The surface physical temperature is T. s The Planck blackbody radiation at time TCF is the topographic configuration factor.

4. The high spatial resolution mountainous area surface temperature hybrid remote sensing inversion algorithm according to claim 3, characterized in that, In S2, the calculation expression for the total thermal radiation of pixels in mountainous areas is: ； In the formula, It is atmospheric upward radiation. It is atmospheric downward radiation, τ dir It is the direct atmospheric transmittance, τ dif It is atmospheric diffuse transmittance. This is atmospheric diffuse radiation, BT m It is the brightness temperature at the top of the atmosphere, B(T) s The surface physical temperature is T. s Planck blackbody radiation at time, SVF is the sky visibility factor, B Let represent the Planck function.

5. The high spatial resolution mountainous area surface temperature hybrid remote sensing inversion algorithm according to claim 1, characterized in that, In S3, the atmospheric profile data consists of 98 lines, and the thermal infrared bands are the five thermal infrared bands of the ASTER sensor.

6. The high spatial resolution mountainous area surface temperature hybrid remote sensing inversion algorithm according to claim 1, characterized in that, In step S4, after obtaining the improved split-window I-SW algorithm, a lookup table for split-window coefficients in mountainous areas is constructed. The expression for the improved split-window I-SW algorithm is: In the formula, The value represents the ground-based radiation brightness temperature in the thermal infrared band. SSP and SVF are the small-scale self-heating parameter and the sky-view factor, respectively. A0~A4 are the regression coefficients of the I-SW algorithm. The first regression coefficient, The second regression coefficient, The third regression coefficient, The fourth regression coefficient, The fifth regression coefficient, To observe the zenith angle, It is the on-board brightness temperature in the thermal infrared band i. It is the on-board brightness temperature in the thermal infrared band J.

7. The high spatial resolution mountainous area surface temperature hybrid remote sensing inversion algorithm according to claim 1, characterized in that, In step S5, the empirical relationship between the minimum emissivity and the difference between the minimum and maximum emissivity of the temperature-emissivity separation algorithm is recalibrated to obtain the ε of the temperature-emissivity separation algorithm for mountainous areas. min In the MMD empirical relation, each SSP corresponds to the ε of the separation algorithm. min - Regression coefficients of the empirical relationship between MMD; where SSP ranges from 0.0 to 1.0 with a step size of 0.

1.

8. The high spatial resolution mountainous area surface temperature hybrid remote sensing inversion algorithm according to claim 1, characterized in that, In S6, the expression for the mountainous surface temperature MLST is: ； In the formula, For band b max The inverse of Planck's function, b max ε represents the band corresponding to the maximum emissivity in the thermal infrared band. max R is the maximum transmit rate for all channels. terrain，bmax b max The contribution of reflected thermal radiation from adjacent terrain, It is band b max Atmospheric downward radiation, T g，bmax This indicates the ground-level radiation temperature corresponding to the band with the highest emissivity in the thermal infrared band.

9. The high spatial resolution mountainous area surface temperature hybrid remote sensing inversion algorithm according to claim 1, characterized in that, In S6, the iterative loop method for inverting the mountain surface temperature (MLST) and effective reflectance (MLSE) of the mountain area using the mountain temperature-emissivity (TES) algorithm is as follows: the iteration stops when the output difference of the surface temperature of the continuous mountain area is less than 0.1 K or when 12 iterations are completed.

Images